Development of Cluster Initialization Methods

Abstract

newline Data, information and knowledge have an important role in various human activities. By processing data, information is extracted, and by analyzing data and information, knowledge is extracted. In today s world, problem of storing, managing and analyzing the huge volumes of data, which is generated regularly by the various sources like sensors, social networking websites, World Wide Web (WWW), Geographical Information System (GIS), automated data capturing applications, etc., has arisen. The problem of storing, managing and analyzing these massive volumes of data leads to the need for large data repositories such as data warehouse, in which data can be stored in a well-structured manner as well as to help in analyzing the stored data with help of OLAP and data mining. To take a holistic view of the research trends in the fields of data warehousing and da ta mining, systematic comprehensive surveys are presented in the thesis. These surveys also presents the current research issues and challenges in the area of data warehousing and data mining for future directions. Clustering is a data processing technique that is extensively used to find novel pat terns in data in the field of data mining and also in classification techniques. It is widely used in numerous applications. Clustering is a descriptive technique based on unsupervised learn ing. It plays a vital role in the classification of data objects into groups (or classes) where the number of groups and their labels are unknown. Therefore, clustering may also consider as initial step of classification to identify the optimal number of groups and to describe their characteristics. Based on the described characteristics, the identified classes can be labeled easily and further, classification can be applied. A number of clustering algorithms are pre sented in the literature. Each clustering algorithm has its application areas, strengths and weaknesses. Each algorithm possesses some strengths and weaknesses. Therefore, a set of clustering algorithms a